Segmentation, Labelling and Speech Corpora 3
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
A Recursive Algorithm for the Forced Alignment of Very Long Audio Segments
Authors:
Pedro J. Moreno, Compaq Computer Corporation (USA)
Chris Joerg, Compaq Computer Corporation (USA)
Jean-Manuel Van Thong, Compaq Computer Corporation (USA)
Oren Glickman, Compaq Computer Corporation (USA)
Page (NA) Paper number 68
Abstract:
In this paper we address the problem of aligning very long (often more than one hour) audio files to their corresponding textual transcripts in an effective manner. We present an efficient recursive technique to solve this problem that works well even on noisy speech signals. The key idea of this algorithm is to turn the forced alignment problem into a recursive speech recognition problem with a gradually restricting dictionary and language model. The algorithm is tolerant to acoustic noise and errors or gaps in the text transcript or audio tracks. We report experimental results on a 3 hour audio file containing TV and radio broadcasts. We will show accurate alignments on speech under a variety of real acoustic conditions such as speech over music and speech over telephone lines. We also report results when the same audio stream has been corrupted with white additive noise or compressed using a popular web encoding format such as RealAudio. This algorithm has been used in our internal multimedia indexing project. It has processed more than 200 hours of audio from varied sources, such as WGBH NOVA documentaries and NPR web audio files. The system aligns speech media content in about one to five times realtime, depending on the acoustic conditions of the audio signal.
SL980068.PDF (From Author)
SL980068.PDF (Rasterized)
TOP
The Selection of Pronunciation Variants: Comparing the Performance of Man and Machine
Authors:
Judith M. Kessens, A2RT, University of Nijmegen (The Netherlands)
Mirjam Wester, A2RT, University of Nijmegen (The Netherlands)
Catia Cucchiarini, A2RT, University of Nijmegen (The Netherlands)
Helmer Strik, A2RT, University of Nijmegen (The Netherlands)
Page (NA) Paper number 372
Abstract:
In this paper the performance of an automatic transcription tool is evaluated. The transcription tool is a Continuous Speech Recognizer (CSR) running in forced recognition mode. For evaluation the performance of the CSR was compared to that of nine expert listeners. Both man and the machine carried out exactly the same task: deciding whether a segment was present or not in 467 cases. It turned out that the performance of the CSR is comparable to that of the experts.
SL980372.PDF (From Author)
SL980372.PDF (Rasterized)
TOP
Acoustic Confidence Measures for Segmenting Broadcast News
Authors:
Jon Barker, University of Sheffield (U.K.)
Gethin Williams, University of Sheffield (U.K.)
Steve Renals, University of Sheffield (U.K.)
Page (NA) Paper number 643
Abstract:
In this paper we define an acoustic confidence measure based on the estimates of local posterior probabilities produced by a HMM/ANN large vocabulary continuous speech recognition system. We use this measure to segment continuous audio into regions where it is and is not appropriate to expend recognition effort. The segmentation is computationally inexpensive and provides reductions in both overall word error rate and decoding time. The technique is evaluated using material from the Broadcast News corpus.
SL980643.PDF (From Author)
SL980643.PDF (Rasterized)
TOP
A Duration-Based Confidence Measure for Automatic Segmentation of Noise Corrupted Speech
Authors:
Bryan L. Pellom, Duke University (USA)
John H.L. Hansen, Duke University (USA)
Page (NA) Paper number 853
Abstract:
In this study, a duration-based measure is formulated for assigning confidence scores to phonetic time-alignments produced by an automatic speech segmentation system. For speech corrupted by additive noise or telephone channel environments, the proposed confidence measure is shown to provide a reliable means by which gross segmentation errors can be automatically detected and marked for human hand correction. The measure is evaluated by computing Receiver Operating Characteristic (ROC) curves to illustrate the expected trade-off in probability of detecting gross segmentation errors versus false alarm rates.
SL980853.PDF (From Author)
TOP
Segmentation and Classification of Broadcast News Audio
Authors:
Thomas Hain, Cambridge University (U.K.)
Philip C. Woodland, Cambridge University (U.K.)
Page (NA) Paper number 851
Abstract:
Broadcast news audio data contains a wide variety of different speakers and audio conditions (channel and background noise). This paper describes a segmentation, gender detection and audio classification scheme for such data which aims to provide a speech recogniser with a stream of reasonably-sized segments, each from a single speaker and audio type while discarding non-speech data. Each segment is labelled as either narrow or wide band and from either a female or male speaker. The segmentation system has been evaluated on the DARPA 1997 broadcast news data set and detailed segmentation accuracy results are presented. It is shown that the speech recognition accuracy for these automatically derived segments is very nearly the same as that for manually segmented data.
SL980851.PDF (From Author)
SL980851.PDF (Rasterized)
TOP
Speaker Recruitment Methods And Speaker Coverage - Experiences From A Large Multilingual Speech Database Collection
Authors:
Børge Lindberg, CPK, Aalborg University (Denmark)
Robrecht Comeyne, Lernout & Hauspie Speech Products NV (Belgium)
Christoph Draxler, IPSK, University of Munich (Germany)
Francesco Senia, CSELT (Italy)
Page (NA) Paper number 1126
Abstract:
With the globalisation and evolving technology of voice-driven man-machine interfaces there is a growing demand for acquisition of spoken language resources in a number of speaker populations being representative for a number of languages and countries. In this paper experience from work within a large consortium in creating large multilingual speech databases for tele-services are reported. In particular the methods and experiences in recruiting speakers for such recordings are reported across a number of participating partners. The reporting is from the SpeechDat project (http://speechdat.phonetik.uni-muenchen.de).